Professional Documents
Culture Documents
JAGDISH MEHRA#
Znstituts Znternationaux de Physique et de Chimie,
1050 Brussels, Belgium
Synopsis
In three papers (1902-1904), Albert Einstein developed the essential principles of the statistical
mechanical approach to thermodynamics before he was twenty-five years old. Einstein’s work
was inspired by Ludwig Boltzmann’s Vorlesungen iiber Gastheorie, but it was completely indepen-
dent of J. Williard Gibbs’s (whose monograph on statistical mechanics was published in 1902).
In certain respects, especially in the search for the existence of atoms of matter and quanta of
radiation by a study of fluctuations, Einstein’s work went beyond Boltzmann’s and Gibbs’s.
For Einstein, his study of the principles of statistical thermodynamics was merely a prelude to
his achievements in quantum theory and brownian motion.
1. Introduction. Albert Einstein, at the age of 23, completed his first paper on
the statistical foundation of thermodynamics in Berne, Switzerland. It was entitled,
“Kinetische Theorie des Warmegleichgewichtes und des zweiten Hauptsatzes der
Thermodynamik” (Kinetic Theory of Thermal Equilibrium and the Second Law
of Thermodynamics). It was received by the Annalen der Physik on 26 June 1902,
and was published a few weeks laterl). This was Einstein’s third scientific paper
to be published”.
The period during which this paper was written had been very critical in Ein-
stein’s life. Since the completion of his “Diplom” studies at the Zurich Polytechnic
(the E.T.H.) in the fall of 1900, he had been constantly in search of a suitable job.
When Einstein started work on the foundations of statistical thermodynamics,
447
448 J. MEHRA
after he had completed his second paper* concerning the molecular interactions,
he was looking forward to his first emp!oyment as a teacher in the technical school
(Technikum) in Winterthur. He held this job only from May to July, 1901.
Starting 15 September 1901, Einstein accepted a position as a junior teacher in a
boys’ school in Schaffhausen, and it was from there that the applied to the Swiss
Patent Office in Bern on 18 December 1901’).
At the end of April 1902 Einstein moved to Bern, where he took up his post
as a Technical Expert, Third Class, on 23 June 1902. Before joining this position
he earned his living by giving private lessons in physics and mathematics. In Bern
he met again Paul Winteler, his old friend from Aarau, who later on married
Einstein’s sister. Einstein made the acquaintance of Maurice Solovine with whom.
and Conrad Habicht, he started the Olympia Academy, their private discussion
group. Also on arrival in Bern he completed, just about the time he joined the
Patent Office, his first paper on the foundations of statistical mechanics, which
was published in the same year (1902) as J. Willard Gibbs’s Elementary Principles
in Statistical Mechanic.rh).
* The essential ideas for his second paper occurred to Einstein early in 1902. In a letter to his
friend Marcel Grossmann, dated 14 April (1902) Einstein wrote: “I am now sure that my theory
of the attractive forces between atoms can be extended to gases and that the characteristic con-
stants of nearly all elements can be specified without undue difficulty.” (This letter is quoted by
Carl Seelig, in ref. 4, pp. 61-62. He assumes that this letter had been written in 1901, probably
because Einstein referred in it to his search for a position as an assistant at a university. How-
ever, he also referred to the help which he had received from Grossmann’s father for the position
at the Patent Office in Bern. Einstein had not applied for this position until December 1901.
In his letter of application, Einstein mentioned that he had seen an advertisement in the “Bundes-
blatt” of 11 December 1901. Seelig believes, therefore, that the letter to Grossmann was from
14 April 1902.)
In his letter of application to the Swiss Patent Office, dated 18 December 1901, Einstein also
noted, “In the first two months [September to November 19011 of my work at [Schaffhausen]
I wrote my doctoral thesis on a topic of the kinetic theory of gases, which I have submitted a
month ago to the Second Section of the Philosophical Faculty of the University of Zurich.”
This letter is reproduced in ref. 5, p. 41.
Thus Einstein had been thinking about the problems of the kinetic theory of gases since early
1901, but he completed his first paper on this subject only in June 1902.
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 449
close to this goal”*. Einstein declared that this goal was to fill this gap. He set
out to obtain the laws of thermal equilibrium, “an extension of the Second Law,
which is of importance for the application of thermodynamics”, and the “mathe-
matical expression for the entropy” by the sole use of “mechanical equations and
the theory of probability”.
In his first paper’) on the statistical foundations of thermodynamics, Einstein
based himself on the model which had been considered by Boltzmann, i.e., systems
with a large number of degrees of freedom satisfying the canonical equations of
dynamics. He introduced the set of systems from which the thermodynamical
properties could be derived, thereby defining the temperature and obtaining the
laws of thermal equilibrium. In his second paper’) dated January 1903, he sought
to develop more general postulates for physical systems satisfying individually
arbitrary linear equations, and to extend Liouville’s theorem to a system of
differential equations more general than the canonical equations of dynamics.
The organization of these two papers was very similar. In a third paper”), dated
Bern, 27 March 1904, Einstein returned to the definition of entropy, deriving an
expression for it analogous to that of Boltzmann and Planck. On the basis of these
considerations he discussed the formulation of the Second Law of Thermo-
dynamics. These three papers represent the highlight of Einstein’s contribution to
the statistical foundation of thermodynamics.
E = L + Va + Vi(t)- (1)
* Boltzmann believed that in the probabilistic interpretation of the entropy he had found a
very general principle which did not have to be limited to the case of gases, and he had expressed
this conviction in his great paper of 1877 (ref. 8). In his Vorlesungen iiber Gastheorie (ref. 7),
however, Boltzmann distinguished between the mathematical significance of the function H (the
measure of the probability of states whose variations were obtained by the detailed analysis of
collisions between the molecules) and its physical significance in relation to the entropy. Boltz-
mann discussed this physical significance explicitly only for special cases, “so as not to make it
incomprehensible with difficulties on account of too much generality.”
In his papers, Einstein explicitly cited Boltzmann’s lectures on gas theory, and it is indeed
possible that the studies of the young Einstein were limited to these.
# It is by considering the external potential V, that Einstein proceeded to generalize the
second law of thermodynamics. In order to realize semipermeable walls, one may assume specific
external forces acting on the molecules, which can be derived from a potential. The potential V
does not enter the Lagrange equation for the momentum q,,, i.e., 9, = Z/2 (dp,!dt).
450 J. MEHRA
where dN denotes the number of systems whose state variables belong to the
domain g at a certain time; g being an infinitesimal domain of all the state variables
Pl, . . ..pn. 413 . ..1 q,,, which is chosen such that E (p, , . .., q,,) lies between E
and E + 6E. Einstein used Liouville’s theorem which states that y (pI, . . ., q,J
does not change with the time if one transforms the state variab!es 1~~) . . . , qII
appropriately to P, , . . . , Qn and the region g to G#. Thus,
The set of systems in which one can relate the distribution function y to the
energy in the specific form (4) is called a “canonical ensemble”*. Thus,
h = + w’(E)/w(E), (7)
where w(E) = J”dL7,, . . . . dx,,, and w’(E) = [dw (E)/dE] co(E) and o’(E) are
always positive, hence h is always a positive quantity.
Jf h is chosen in this way, then the integral in eq. (6) reduces to a quantity
independent of 7, and the number of systems whose variables p1 , . . . , qn lie within
the specified limits assumes the expression
Eq. (8) also represents the probability (if suitably normalized by N) that the state
variables of a thermometer G, connected with a heat reservoir 2, lie between
infinitesimally close limits if the state has become stationary.
The state of the thermometer is defined by eq. (8); 12represents the temperature
function, and every observable of the thermometer system CJsimply depends on it
* The name “canonical ensemble” was proposed by Gibbs, ref. 6, chs. IVff. Einstein gave
no further justification for his procedure, and did not enlarge upon the difference between
microcanonical and canonical ensembles as Gibbs did in his book.
4 Eq. (7) is derived by noting that the integral appearing in eq. (6) is a function X of (E - 7)
and 2 (E - q) z x(E) - E;c’ (E) = x(E) if x’(E) = 0. Since x(E) = eeZhE w(E), eq. (7)
follows.
The positivity of h is established with the help of a mathematical trick. For n > 2 and LY> 1
one proves that
and from this it follows that w’(E) z 0, since o(E) can be composed of the differentials dw (E)
which satisfy the above inequality such that do) (Ez) > do) (Er) if Ez > El.
452 J. MEHRA
and
The quantities r were called the “momentoids” by Bolttmann. The average kinetic
energy per degree of freedom, or corresponding to a momentoid, from eq. (8) is
given by,
17 = a(Vi+V,-L) +d
Y (11)
ab
The work done on the system by the forces nV in time dt represents the amount
of heat dQ transferred to the system,
* This step involves an assumption which Einstein did not prove in his first paper, ref. 1.
It is that the “time average” of a system is related to the average over a microcanonical ensemble.
Einstein returned to the consideration of the time ensemble in his subsequent paper (ref. 8), but
there also he did not give a specific proof of the assumption involved. This assumption has to
do with Boltzmann’s “ergodic” hypothesis. This fact seemed to be so obvious to the young
454 J. MEHRA
stein then considered a very large number of stationary systems with identical
values of I7 and V, (or, rather, h lying between /z and iz + 611, V, between P’:, and
V, + SV,, and V between V and Y + SV). The number of systems whose state
variables lie within an infinitesimal domain g is then given by
For given lz and V,, I/ can be chosen such that the constant A becomes unity:
and the new total potential is expressed as V*. Then.
441
i3v*
c -dp,=!$
8P”
...
dq, V:k 0 (e-2h(“*+L)).
This expression denotes the increase in the average potential energy of the system
that would occur if the state distribution were to change in accordance with S V*
and Bh, without an explicit change in V itself. Thus,
assuming that the total number of systems considered does not change, i: and i
being the average values of the potential and kinetic energies, respectively.
From eq. (9), h = n/4L, hence 6/z = -(n/4L’) 8L. From eqs. (lo), (13) and (15),
one thus obtains,
Einstein (whose knowledge of the kinetic theory was derived almost completely from Boltzmann’s
“Gastheorie”) that he did not find it necessary to go into further details. Boltzmann’s “ergodic”
hypothesis was first examined in detail by P. and T.Ehrenfest in their article on “The Conceptual
Foundations of Statistical Mechanics” (ref. 1I).
* Ref. 1, p. 432.
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 455
In eq. (17), only the first term is of physical interest, since the second term represents
a constant and the third does not depend either on E or T.
Einstein concluded, “The expression for the entropy E is noteworthy because
it depends on E and T alone, and the special form of E as the sum of potential
and kinetic energies does not enter. This fact allows us to presume that our results
are more genera1 than the mechanical formulation employed [by us], especially
since the expression for 12[eq. (IO)] exhibits the same property.” *
“Concerning the nature of the forces which correspond to the potential V,“,
Einstein continued, “nothing need be assumed, not even whether these forces exist
in nature. The mechanical theory of heat requires that we should arrive at the
correct results if we apply Carnot’s principle to idea1 processes that can be gener-
ated from the observed ones by the introduction of arbitrary V,. Naturally the
results obtained from the theoretical consideration of these processes have real
significance only when the arbitrary, auxiliary forces V, do no appear in them” ‘.
* Ref. 1, p. 433.
* Ref. I, p. 433. By means of such a treatment one can verify the application of the second
law of thermodynamics to processes in which one assumes, for instance, the existence of semi-
permeable walls which can be realized physically only with difficulty.
: Ref. 4, p. 71.
* Ref. 4, p. 69.
456 J. MEHRA
isolated systems never pass into those which are less probable, there follows the
second law of thermodynamics in its most general form, namely the impossibility
of a pevpetuum mobile of the second kind”r2).
What caused Einstein such satisfaction, and which improvements and additions
had he made on the substance of his previous paper on the mechanical explanation
of the second law of thermodynamics?
In his second paper Einstein attacked the problem of deriving the laws of
thermodynamics from statistical considerations more general than those used in
kinetic theory, which he had employed in his first paper.
dPi
- = pli (PI > ...- PA. (18)
dt
where the vr are given, unique functions. There exists at least one integral (dis-
regarding the others) of eqs. (18) which does not include the time, namely the
energy.
In order to define a stationary state, Einstein used the property that the variables
PI 3 ..-> p,, of the system take on the same values again and again with the same
multiplicity. The same fact can be expressed as follows: Given a domain r of the
phase space (of state variables), then the ratio of the time during which the state
lies in that domain, to the total time T, tends to a constant value in the limit,
which is independent of the time. This ratio is infinitesimally small for an in-
finitesimal domain of the phase space. At any arbitrary instant, the number of
systems whose state variables lie within the domain r is given by Nlim,,, t/T,
which does not depend on the arbitrarily chosen instant.
The number of systems whose state variables occupy at any time an infinitesimal
domain r of the coordinates pl, . . . , pn is given by
’ a (a%) dp,
dN,+,, = dN, - c ~ ... dp, = dN,. (22)
v=l ap,
From this, it immediately follows that
+ y(E)
1
= const emm. (23)
Einstein then defined the new state variables 17, , . . . ,17,, and absorbed the factor
e-“’ in one of the coordinates, obtaining
This equation is identical with eq. (3) derived in his first paper, but the spirit of
the new derivation was completely different.
3.2. Heat equilibrium. Having obtained eq. (24) Einstein again considered
a microcanonical ensemble of systems consisting of a heat reservoir 2 (with state
variables 17, , . . . , UL) and a thermometer (T (with state variables z1 , . . . , nJ. The
energy E, constant to within dE, of a system Z + ~7can be divided into two parts,
E=H+rj, (25)
with a positive function h of the energy, which thus determines the state of the
system rs. The absolute temperature of the whole system is defined by h which,
as before, is given by
T = 1/4hx. (27)
then the definition of temperature in eq. (27) coincides with that of the kinetic
theory.
3.3. The second law and the entropy. In his first paper Einstein had
employed explicitly the lagrangian formulation in his derivation of the second law
[eqs. (ll)-(16)]. He again wished to obtain the most general formulation of the
second law, but without specifically using the equations of mechanics. In order
to proceed, several steps were necessary: (a) definition of infinitely slow processes;
(b) introduction of the entropy; (c)introduction of the probability of a distribution;
(d) derivation of the second law.
(28)
(29)
(30)
* Both of these processes are used in thermodynamics. Boltzmann had made extensive use of
processes in which the volume remains constant, i.e., isopycnic processes.
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 459
where the first term comes from the adiabatic change (L’s being the parameters
by which system 2 influences the system undergoing the change), and is zero for
an isopycnic change; and the second term arises from an isopycnic change [eq. (29)],
being zero for an adiabatic process.
Before and after the process the state of the system, which has been in contact
with a large system (reservoir) during the process, should be stationary. Its state
may be described by the probability of a canonical ensemble [as in eq. (8)] as
From eq. (33), it follows that the term within the parentheses must vanish, i.e.
dC - 2Edh - ?hxzd/! = 0.
Comparing the expression for dQ in eqs. (29) and (30), and using eq. (34), one
obtains 2h dQ = d (2hE - C), and with 1/4h = XT,
up to a constant.
If the domain g is split into 1 parts such that jg, = jy2= ... jg,,then the probabil-
ities W, , . . . , W,, would also be equal: i.e.,
N
N!
w=
0 f
F1 ! s,! .‘. FI!
, (38)
logW=const.-$_elogf:=const-Jalog&dpr...dp,,, (39)
E=l
which follows from eq. (39) and the fact that the number of systems N remains
constant.
dw, = eC(~)-2h(l)E(~)
jdp:” . . . dpi”,
cl
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 461
where the indices pertain to the subsystem IJ~. The probability of the distribution
of the total system among the state variables pl, . . . , p,, is given by a product of
the probabilities of distribution of the subsystems, i.e.,
since the subsystems are assumed to be independent of each other. If there are N
such total systems, then their distribution function, defined by eq. (20) is given by
That is, the sum of the entropies of the subsystems of an isolated system after any
arbitrary process is equal to or greater than the sum of the entropies of the sub-
systems before the process.
In order to complete the proof of the impossibility of a perpetuum mobiZe of the
second kind, Einstein considered an isolated system consisting of a heat reservoir
2’ with an infinitely large energy, an engine M, and subsystems L’, , 2Y2, . . . in
adiabatic contact with each other, all the subsystems being in a stationary state.
The energies of the subsystems ,X1, Z2, . . , are infinitely large compared to the
energy of the engine M.
Now the engine M goes through an arbitrary cyclic process during which the
state distributions of the systems Z, , .X2, . . . are changed adiabatically infinitely
slowly; i.e., the engine does the work and extracts an amount of heat Q from the
reservoir L’. At the end of the process the mutual adiabatic interaction of Z:, , Z; , . . .
will be different from that at the beginning of the process. The engine M has thus
converted heat Q into work.
What is the increase in entropy of the individual systems during this process?
The increase in the entropy of the heat reservoir, according to earlier discussion,
is -Q/T, where Tis the absolute temperature. The entropy of the systemsx, , Z2, . . .
does not change at all during the process since they undergo only an infinitely
slow adiabatic effect. The increase in the entropy S’ - S of the total system is
thus given by,
S’ - S = -Q/T. (44)
462 J. MEHRA
According to the conclusion obtained earlier, this quantity is always greater than
or equal to zero, and it follows that
Q 5 0. (45)
Eq. (45) expresses the impossibility of the existence of a perpetual motion machine
of the second kind.
4. The general molecular theory of heat. Einstein completed his third paper on
the statistical foundation of thermodynamics on 27 March 1904lO). The circum-
stances in which he1 ived were now quite pleasant. His position at the Patent
Office would become permanent later in the year*. He had gathered a circle of
friends, and Michele Besso, who would serve as Einstein’s sounding board for
new ideas for half a century, also came to Bern in 1904 and joined the Patent
Office.
In his third paper Einstein supplemented his earlier work on the statistical
foundation of thermodynamics, discussing five topics.
(i) He derived an expression for the entropy, which was analogous to Boltz-
mann’s expression obtained from the theory of ideal gases and to Planck’s expres-
sion used in the theory of heat radiation.
(ii) Einstein gave a condensed derivation of the second law of thermodynamics,
which extensively employed the concept of probability, as defined in eq. (39),
developed in his previous paper.
(iii) He discussed the meaning of the constant x (the Boltzmann constant with
a factor +) in the kinetic theory of gases. Einstein presented a numerical value for
this constant, derived from the relation between the gas constant R, and N, the
number of atoms in a gram atom.
(iv) In an attempt to obtain the general significance of the constant ;c, Einstein
discovered for himself the concept of fluctuations: it determines the fluctuation of
a thermodynamical system.
(v) Einstein was not satisfied with the fact that fluctuations in a normal thermo-
dynamical system (with a large number of degrees of freedom) are unobservably
small, a fact with which Boltzmann and Gibbs seemed to have been satisfied. In
searching for a system in which the energy fluctuations might become large, Ein-
stein turned his attention to the phenomenon of heat radiation in empty space.
In a region the dimensions of which are of the order of the wavelength of radiation,
Einstein expected energy fluctuations of the order of magnitude of the energy
itself.
In order to prove this idea, Einstein calculated the wavelength ;i,,, of the heat-
radiation spectrum with maximum energy. The energy of a cavity of volume v
and temperature T is given by the Stefan-Boltzmann law. Einstein related this
4.1. The second law of thermodynamics. Withthe help of eqs. (36) and
(7), Einstein derived
Now he turned his attention to a new derivation of the second law. A system which
has been in contact for a very long time with a heat reservoir of temperature To
absorbs an energy between E and E + dE with the probability [eq. (31)]:
The reservoirs are connected with an engine which performs a cyclic process, but
does not exchange energy with them. This means that the energy and entropy are
exchanged only between the I reservoirs, and since the total energy remains con-
stant just the entropy term in eq. (48) is relevant for further discussion.
Introducing once again the assumption that states with lesser probability never
follow states with higher probability, or
ZY 2 TO!;, (49)
where the prime denotes the later state, Einstein concluded that
4.2. The constant it and fluctuations. Given the distribution, eq. (47)
the average kinetic energy of an atom in a system of atoms is given by
where x1, . . .) z, are the position coordinates, and E, , . . . , <,, the momentum co-
ordinates, of the y1atoms, and #J (x,, . . . , z,,j is the potential energy of the atomic
system.
Now from the kinetic theory of gases Einstein deduced that, since pc = RT,
and
with R, the gas constant, equal to 8.31 x IO’, and N = 6.4 x 10Z3.
Einstein was not satisfied with this determination of it, because it did not
essentially go beyond what Boltzmann had done*. Given the assumption of the
molecular kinetic theory of matter, Einstein thought that he had to confer a deeper
significance on the fundamental constant x. He thereby discovered independently
a result which had by then already been included in Gibbs’s book# ; he had derived
the fluctuation equation.
For this purpose, Einstein studied a system which is connected with a heat
reservoir of temperature T. The probability for this system to take the energy E
is obtained from eq. (47), and the average energy I? can be calculated to be
Eq. (55) merely indicates the fact that the average value of the quantity in the
parentheses vanishes; hence
* In fact, Boltzmann did not use the constant x, nor did he give a number for it. This was
done for the first time by Planck in 1899 [M. Planck, iiber irreversible Strahlungsvorglnge,
s. Mitteilung, S.-B. Preuss. Akad. Wisr., 440-480, 18991.
i: Ref. 6, ch. 7, eq. (200).
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 465
1 cz E= z E=, (57)
i? = CuT4, (58)
where C = 7.06 x lo-l5 (cgs units) 14, ls). He obtained, using the value from
eq. (52),
2 (X/C)“” 0.42
il “IaXZ:u 113 z =-*
T T
* In pursuing the consequences of eq. (56) and the physical consequences of the quantity 2,
Einstein went beyond Boltzmann and Gibbs. Both Boltzmann and Gibbs had emphasized that
in large systems, which one has to consider, the deviation from the equilibrium value is immeasur-
ably small as soon as the system has attained equilibrium.
Thus Boltzmann had remarked: “Even in the smallest neighbourhood of the tiniest particles
suspended in a gas, the number of molecules is already so large that it seems futile to hope for
any observable deviation, even in a very small time, from the limits that the phenomena would
approach in the case of an infinite number of molecules.” (Ref. 7, English translation, p. 318.)
Gibbs had noted: “It follows that to human experience and observation with respect to such
an ensemble as we are considering . [the fluctuations] would be in general vanishing quantities,
since such experience would not be wide enough to embrace the more considerable divergencies
from the mean values, and such observation not nice enough to distinguish the ordinary
divergencies. In other words, such ensembles would appear to human observation as ensembles
of systems of uniform energy, and in which the potential and kinetic energies (supposing that
there were means of measuring these quantities separately) had each separately uniform values”.
(Ref. 6, pp. 74-75.)
Einstein, being younger and less cautious than both Boltzmann and Gibbs, and not having
their wisdom and experience, turned his attention immediately to a very controversial field of
his time, the theory of heat radiation. It is remarkable that Boltzmann, several years earlier, had
struggled with Planck and denied the fact that one could apply the concept of entropy to radiation
at all (see ref. 13). Einstein, fortunately, had no such scruples.
466 J. MEHRA
The experimental value of the constant in Wien’s displacement law was known
to be,
Both the temperature dependence and the order of magnitude of the factor in
Wien’s law came out correctly from Einstein’s general theory. Einstein concluded,
“I believe this accord cannot be attributed to chance because of the general nature
of our assumption”*.
* Ref. 10, p. 362. One cannot, however, take this numerical agreement too seriously, because
x appears as a cube-root in eq. (59) and, therefore, A,,, does not depend very critically on 1~.
However, for Einstein, this success became the starting point of the ideas which led to his paper
on the light quantum in 1905.
* The phrase “the rational foundation of thermodynamics” was used by J. W.Gibbs as the
subtitle of his monograph on the “Elementary Principles in Statistical Mechanics”, ref. 6.
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 467
atoms. Boltzmann began a series of new and fundamental studies in 1871 with a
paper on “the heat equilibrium between polyatomic gas molecules”, in which he
introduced his theory of time ensembles22), and gave in a following paper “an
analytical proof of the second law” based on the law of equilibrium of kinetic
energy. We shall return to the latter paper, because the “proof” of the second law
that Einstein gave was very similar to Boltzmann’s.
Boltzmann continued his studies on the thermal equilibrium of gases, elaborating
the “Stosszahlansatz”, and introducing the famous function H (which he denoted
by E) related to the entropy in 187224). There he proved that the distribution
“after a very long time must tend to the maxwellian” (see ref. 24 and ref. 19, vol. I,
p. 345), and also introduced his probability method for the first time, which he
would use frequently later on (as would Planck for the derivation of his radiation
formula) : the replacement of the entropy integral by a sum. Boltzmann also stated
there his theory of time ensembles more clearly.
Boltzmann returned to the statistical mechanics of gases in 1875”* ‘h). He
considered various problems of the mechanical theory of heat in 1877, and an-
swered Loschmidt’s objections concerning his mechanical derivation of the second
law, the so-called “reversibility paradox” [“Umkehreinwand”, as it was called by
Ehrenfest]27). Also in 1877, Boltzmann again discussed the relation between the
second law, the theory of probability, and the laws of thermal equilibrium2*).
There he stated clearly the proportionality between entropy and the “measure of
permutability”29).
In a review of Maxwell’s paper, “On Boltzmann’s theorem concerning the
average distribution of the kinetic energy in a system of material points”3g),
Boltzmann made a clear distinction between his own point of view and Maxwell’s :
“There is a difference in the points of view of Maxwell and Boltzmann, in so far
as the former characterizes the probability of a state by the time during which the
system occupies this state on the average, whereas the latter assumes infinitely
many identical systems with all possible initial states. The proportion of the number
of systems which have a [particular] state in relation to the total number of systems
determines the probability in question”3 ‘).
During the following years many applications of Boltzmann’s ideas can be
found in his papers. He returned to fundamental questions again in 1887,
supplementing the proof of the second law which he had given in 187132* 33).
Another study, which is particularly important in the present context, was Boltz-
mann’s paper with G. H. Bryan in 1894 entitled, “On the mechanical analogy for
the thermal equilibrium of two bodies in contact”34).
In three papers during 1896-1897 Boltzmann took up the questions raised by
E.Zermelo, i.e., can one really derive the second law of thermodynamics from a
mechanics which is fully time invariant? Boltzmann noted that the singular initial
states which do not approach the Maxwell distribution are very small in number
compared to those that do. Consequently there is no difficulty in explaining
468 J. MEHRA
* Ernest Zermelo was a German mathematician who had been a pupil of Max Planck. He
wrote very sharply against Boltzmann. Professor G.E.Uhlenbeck once told me that he had
heard from Ehrenfest that Boltzmann always referred to Zermelo as “Dieser Halunke” (rogue
or villain).
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 469
of atoms and molecules, and if one could establish correct laws for these then they
would be valid in general. Most of his proofs had been established with gases in
mind, and were, therefore, not general enough for someone as critical and ambitious
as Einstein.
Having studied the works of Mach (ref. 4, pp. 39, 40, 69, 80, 85), Einstein was
aware of the criticism that had been levelled against the molecular-kinetic theory
of matter. Einstein looked for phenomena that would provide incontestable proof
of the existence of molecules, and it was this search that brought him to examine
the statistical foundation of thermodynamics. “My major aim in this was to find
facts which would guarantee as much as possible the existence of atoms of definite
finite size,” as Einstein recalled over four decades later4’). And, in order to pursue
this programme, he did not wish to restrict himself to systems of more or less free
gas mo!ecules.
Einstein borrowed from Boltzmann what he needed, such as the systems of
equal energy which Gibbs called “microcanonical ensembles”. He invented the
time ensemble, because no explicit remarks about it were contained in Boltzmann’s
“Gastheorie”, although Boltzmann had discussed it in his papers*. As for Ein-
stein’s derivation of the second law [ref. 1, our section (2.3)], it also had a close
similarity to Boltzmann’s derivation2”); although Einstein talked about general
systems and Boltzmann about molecules; Einstein’s proof was not more general.
* Boltzmann had introduced the time ensemble explicitly in 1871, ref. 23 (or ref. 19, vol. I,
p. 288): “If we keep the external conditions (temperature, external forces), to which the body is
subject, unchanged during a very long time T, and let t denote a fraction of T during which its
state variables x1, JJ~, .,z,, ol, .... co,, lie between the limits x1 and x1 + dxl, y, and
~1 + dy, . . . . o, and CO, + dw,, then the ratio r/T is the period during which the state of the
body lies on the average between the limits indicated.” In a previous paper (ref. 46, or ref. 19,
vol. I, pp. 259-287, 282), he had calculated this time to be
of a great number of particles, or more precisely, they express the laws of mechanics
for such systems as they appear to beings who have not the fineness of perception
to enable them to appreciate quantities of the order of magnitude of those which
relate to single particles, and who cannot repeat their e,!periments often enough
to obtain any but the most probable results” [ref. 6, p. (viii)].
This means that for systems of many particles or large number of degrees of
freedom, one can only measure an average, or “expectation” value. How does one
relate an observed value to a calculated value? Actually there are two questions
here: (i) Given a macroscopic parameter of a system, which statistical ensemble
should one choose in order to account for it? (ii) How should one define a macro-
scopic measurement?
The problem is, however, not solved if one defines, on an a priori basis, certain
ensembles which are easy to handle. This is exactly what Gibbs did in his book.
He defined the canonical ensemble as a (mathematically) practical tool for handling
questions of equilibrium thermodynamics. Gibbs both treated the microcanonical
ensemble, the system preferred by Boltzmann, and also referred to the time
ensemble*.
Gibbs had written a classic, both in form and content, which attracts the reader
even today for its Dirac-like economy of style, its appeal to rigour and abstraction,
and in its day presented a contrast to the labyrinthine vastness of Boltzmann’s
original memoirs.
Einstein’s attempt to construct a statistical basis for the laws of thermodynamics
did not arouse any interest at the time, and was almost forgotten later on in view
of his other great achievements. Einstein was a young researcher still unknown,
while Gibbs was a celebrated master of thermodynamics and mathematical
physics. Even in 1911, when Einstein was famous for his work on quantum
theory, relativity, and brownian motion, the review4g) of the conceptual foundations
of the statistical approach in mechanics by Paul and Tatiana Ehrenfest made only
a passing reference to Einstein’s work on statistical thermodynamics, considering
it as being “related to Gibbs’s treatment”.
The kinship in the approaches of Gibbs and Einstein was remarkable. Both had
proceeded from the common source of Boltzmann’s achievements, and both
sought to attain the greatest generality. For specific results, Einstein also employed
canonical ensembles. The detailed analysis of molecular collisions had played an
important role in Boltzmann’s work; of this, there was no trace in the work of
Gibbs or Einstein50).
In his work Gibbs renounced all hypotheses dealing with the detailed structure
of matter. “One is building on an insecure foundation, who rests his work on
hypotheses concerning the structure of matter,” he wrote (ref. 6, ch. X). Instead
he limited himself only to the statistical-mechanical interpretation of thermo-
dynamics. In this sense, Einstein’s point of departure and treatment were bolder.
* See ref. 6, ch. XII, Dover edition, in particular p. 142.
472 J. MEHRA
He did not doubt the possibility of obtaining from the statistical method some
proof of the reality of the existence of atoms and molecules. Einstein was also
concerned about associating entropy to probability in practice, thereby giving it
a really statistical character.
Gibbs had had years of preparation for writing a classical treatise. The young,
inexperienced, and inestablished Einstein had packed the essence of the new
statistical foundation of thermodynamics in two short papers’s9) at a difficult time
in his life, working without the benefit of participation in the atmosphere in which
science is normally pursued. Einstein had built upon Boltzmann, surpassing him
in generality at points, and “anticipated” Gibbs. In his third paperlo) of 1904,
Einstein went beyond both Gibbs and Boltzmann in looking for an observable
effect in nature due to fluctuations, thereby seeking to establish the validity of the
molecular-kinetic theory of matter. “I have discovered,” he wrote to his friend
Conrad Habicht on 14 April 1904, “in the simplest possible way the relationship
between the size of the elementary quanta of matter and the wavelengths of
radiation.” (ref. 4, p. 74). This bespoke the originality of Einstein’s insights, of
which he was to give ample proof in short order.
6.3. Paul Hertz’s criticism. In 1910 Paul Hertz, then a “Privatdozent” at
the University of Heidelberg, published two papers in the Annalen der Physik in
which he reviewed the contributions of Gibbs and Einstein to the mechanical
foundations of thermodynamics5’). Hertz himself attempted to derive the second
law of thermodynamics by seeking to generalize the methods of Gibbs and Einstein.
Hertz took issue with Einstein concerning certain points in the treatment of
thermal equilibrium in his first paperr). Einstein had concluded that: (i) if two
bodies have the same temperature before contact, they will also have the same
temperature afterwards; (ii) if two bodies in contact are separated, then both will
have the same temperature as the composite body before. Hertz’s criticism was
that “Einstein tries to make these two laws plausible. However, what he actually
demonstrates cannot even be regarded as a plausibility argument, and in fact
signifies a renunciation of his own basic principles. He seeks to prove the law of
separation [of bodies] by declaring it to be probable that the gradual separation
of a body causes only a small change in its state while it [still] touches a part
[of the other body]. Nevertheless each separation means the creation of a new
mechanism, and it is not evident what connection would exist between the states,
before and after the separation, of a system which is in contact with a part of the
separated body. Similarly, it is not justified to consider the validity of the law of
contact [between bodies] as a logical consequence of the law of separation. because
the processes considered are reversible”*.
* Ref. 51, pp. 247-248. Hertz’s comments are directed to Einstein’s conclusions (iv) and (v)
in section (2.2). It is interesting to note that Boltzmann, in his paper with Bryan (ref. 34), had
used an argument similar to Einstein’s to prove the heat equilibrium between two types of
molecules. Boltzmann had, however, applied his mechanism of collisions to obtain equilibrium.
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 473
Hertz proposed to join two systems of equal temperature, and to prove that the
combined system has the same temperature. For this, in the microcanonical
systems in question, one has only to distribute the energies appropriately. The law
of contact is, therefore, not a mechanical law, whereas the law of separation is.
Hertz also criticized Einstein’s proof of the second law [section (3.3)], in
particular his application of the assumption that more probable states follow less
probable ones, which seemed unnecessary to Hertz in connection with the ensemble
theory used by Einstein. In short, Hertz maintained that if a virtual ensemble were
assumed then there would be two probabilities, one referring to the distribution
of the systems and the other to the distribution of one system in the course of time.
Since Einstein had talked about the separation of systems, the danger was evident
that the entropy would change discontinuously.
Einstein responded to Hertz’s critique both in a personal discussion with him
and in print 52). “His [Hertz’s] remarks concerning my treatment of temperature
equilibrium are based on a misunderstanding, which arose because [my] formula-
tion was brief and not sufficiently careful.” Einstein went on to say, “I should still
mention that, in my view, the path indicated by Gibbs in his book, which consists
in proceeding directly with a canonical ensemble, is to be preferred to the one
proposed by me. Had I been familiar with Gibbs’s book at the time, I would not
at all have published those papers, and would have limited myself to the treatment
of a few points” (ref. 52, p. 176).
Einstein did not know Gibbs’s book at that time, and he has to be regarded,
together with Gibbs, as the founder of the ensemble-theoretic basis of statistical
thermodynamics.
* Tolman introduced the “axiom of equal a priori probabilities”, which says that equally
large areas of the phase space have equal a priori probabilities. This assumption is independent
of all other axioms, and has to be added to the laws of mechanics in order to complete the basis
of statistical mechanics.
414 J. MEHRA
Boltzmann had proposed the ergodic hypothesis: that the system, in the course
of its evolution, goes through each point in phase space consistent with its (con-
served) energy * . Paul and Tatiana Ehrenfest replaced this hypothesis with the
weaker quasiergodic hypothesis : that each trajectory approaches arbitrarily close
to every point of the energy surface in phase space4’). The strong form of the
ergodic hypothesis was disproved by Rosenthal and Plancherels3* 54)#. The first
proof of the quasiergodic hypothesis was attempted by Rosenthal”), but a more
rigorous treatment by Fermi applied only to the so-called “canonical normal
systems”56). However, Von Neumann and Birkhoff proved the equality F”’ = 2
under very general assumptions, without using the quasiergodic hypothesis5’).
All this was far afield from Boltzmann, Gibbs and Einstein’s initial consider-
ations. In 1949 Einstein, referring to his early work, simply noted, “Not acquainted
with the earlier investigations of Boltzmann and Gibbs, which had appeared earlier
and actually exhausted the subject, I developed the statistical mechanics and
molecular-kinetic theory of thermodynamics which was based on the former”47).
This self-education led immediately to Einstein’s work on brownian motion and
the quantum theory, which helped him achieve his “major aim of guaranteeing
as much as possible the existence of atoms of matter” and quanta of radiation.
His work on statistical thermodynamics had merely been a preparation for these
achievements.
In January 1954 Einstein was asked: who were the greatest men, the most
powerful thinkers whom he had known ? He answered without hesitation,
“Lorentz”. Then he added, “I never met Willard Gibbs; perhaps, had I done so,
I might have placed him beside Lorentz”” ). This homage to Gibbs was not based
on any intellectual debt.
REFERENCES
* This strong form of the statement can be found in the papers of both Boltzmann (ref. 21)
and Maxwell (ref. 18).
# The first recognition of the mathematical distinction between ergodic and quasi-ergodic
system is actually due to P. Hertz, ref. 51.
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 475
7) Boltzmann, L., Vorlesungen iiber Gastheorie, J. A.Barth (Leipzig: part I, 1896; part II, 1898).
Reprinted 1910-1912. English translation by S.G.Brush, Lectures on Gas Theory, Univ.
of California Press (Berkeley and Los Angeles, 1964).
8) Boltzmann, L., Wien. Ber. 76 (1877) 373.
9) Einstein, A., Eine Theorie der Grundlagen der Thermodynamik, Ann. Physik 11 (1903)
170-187 [Received 26 January 19031.
10) Einstein, A., Zur allgemeinen molekularen Theorie der Wlrme, Ann. Physik 14 (1904)
354-362 [Completed 17 March 1904, Received 19 March 19041.
11) Ehrenfest, P. and Ehrenfest, T., Encyclopadie der Mathematischen Wissenschaften, vol. IV
2 II, B.G. Teubner (Leipzig, 1912).
12) Albert Einstein, Michele Besso, Correspondence 1903-1955, P. Speziali, ed., Hermann (Paris,
1972) p. 3.
13) Boltzmann, L., Ber. Press. Akad. Wiss. (1897) 660; (1989) 182.
14) Stefan, J., Ber. Kais. Akad. Wiss. (Wien) 79 (1879) 391.
15) Boltzmann, L., Wien. Ann. Phys. 22 (1884) 291.
16) Maxwell, J. C., “Illustrations of the dynamical theory of gases”, Phil. Mag. (4) 19 (1860) 19;
20 (1860) 21, 33.
17) Maxwell, J. C., “On the dynamical theory of gases”, originally published in the Phil. Trans.
Roy, Sot. 157 (1867) 49, and in Phil. Mag. 32 (1866) 390; 35 (1868) 129, 185.
18) Maxwell, J.C., “On Boltzmann’s theorem on the average distribution of energy in a system
of matter points”, Trans. Camb. Phil. Sot. 12 (1879) 547.
19) Wissenschaftliche Abhandlungen von Ludwig Boltzmann, vols. I-III, F.HasenGhrl, ed.,
J.A. Barth (Leipzig, 1909).
20) Boltzmann, L., “Uber die mechanische Bedeutung des zweiten Hauptsatzes der WPrme-
lehre”, Sitz,-Ber. Kais. Akad. Wiss. (Wien) 53 (1866) 195 (see also ref. 19, vol. I, pp. 9-33).
21) Boltzmann, L., “Studien fiber das Gleichgewicht der lebendigen Kraft zwischen bewegten
materiellen Punkten”, Wiener Ber. 58 (1868) 517 (see also ref. 19, vol. I, pp. 49-96).
22) Boltzmann, L., “Uber das Warmegleichgewicht zwischen mehratomigen Gasmolektilen”,
Wiener Ber. 63 (1871) 397 (see also ref. 19, vol. I, pp. 236-258).
23) Boltzmann, L., “Analytischer Beweis des zweiten Hauptsatzes der mechanischen Wlrme-
theorie aus den SPtzen fiber das Gleichgewicht der lebendigen Kraft”, Wiener Ber. 63 (1871)
712 (see also ref. 19, vol. I, pp. 288-308).
24) Boltzmann, L., “Weitere Studien fiber das Warmegleichgewicht unter Gasmolekiilen”,
Wiener Ber. 66 (1872) 275 (see also ref. 19, vol. I, pp. 316-402).
25) Boltzmann, L., “Uber das Warmegleichgewicht von Gasen, auf welche BuBere Krafte wir-
ken”, Wiener Ber. 72 (1875) 427; Wien. Anz. 12 (1875) 174; Phil. Mag. 4 50 (1875) 495
(see also ref. 19, vol. II, pp. l-30).
26) Boltzmann, L., “fiber die Aufstellung und Integration der Gleichungen, welche die
Molekularbewegung in Gasen bestimmen”, Wiener Ber. 74 (1876) 503; Wien. Anz. 13 (1876)
204 (see also ref. 19, vol. II, pp. 55-102).
27) Boltzmann, L., “Bemerkungen tiber einige Probleme der mechanischen Warmetheorie”,
Wiener Ber. 75 (1877) 62 (see also ref. 19, vol. II, pp. 112-148).
28) Boltzmann, L., “Uber die Beziehung zwischen dem zweiten Hauptsatz der mechanischen
Warmetheorie und der Wahrscheinlichkeitsrechnung, respektive den SPtzen iiber das Wlrme-
gleichgewicht”, Wiener Ber. 76 (1877) 373 (see also ref. 19, vol. II, pp. 164-223).
29) See also Boltzmann, L., “Weitere Bemerkungen tiber einige Probleme der mechanischen
Warmetheorie”, Wiener Ber. 78 (1878) 7 (see also ref. 19, vol. II, pp. 250-288).
30) Maxwell, J. C., Trans. Camb. Phil. Sot. 12 (1879) 547.
31) Boltzmann, L., Ann. Physik Beibl. 5 (1881) 403 (see also ref. 19, vol. II, pp. 582-595).
,
476 J. MEHRA
32) Boltzmann, L., “iiber mechanische Analogien des zweiten Hauptsatzes der Thermo-
dynamik”, Crelles Journal 100 (1887) 201 (see also ref. 19, vol. III, pp. 258-271).
33) Boltzmann, L., “iiber einige Fragen der kinetischen Gastheorie”, Wiener Ber. 96 (1887) 891 ;
Phil. Mag. 25 (1888) 81 (see also ref. 19, vol. 111, pp. 293-320).
34) Boltzmann, L. and Bryan, G.H., “Uber die mechanische Analogie des Wlrmegleichgewichtes
zweier sich beruhrender Korper”, Wiener Ber. 103 (1894) 1125; Proc. Phys. Sot. 13 (1895)
48.5 (see also ref. 19, vol. Ill, pp. 510-519).
35) Boltzmann, L. and Bryan, G.H., See also “On the application of the determinental relation
to the kinetic theory of polyatomic gases”,
Rep. Brit. Assoc. Adv. Science (Oxford, 1894)
pp. 102-106 (see also ref. 19, vol. III, pp. 520-525).
36) Boltzmann, L., “Entgegnung auf die warmetheoretischen Betrachtungen des Hrn. E.Zer-
melo”, Annalen der Physik 57 (1896) 773 (see also ref. 19, vol. III, pp. 567-578). English
translation as “Reply to Zermelo’s Remarks on the Theory of Heat” in Brush, S.G.. Kinetic
Theory, vol. 2, Pergamon Press (Oxford, 1966) pp. 218-228.
37) Boltzmann, L., “Zu Hrn. Zermelos Abhandlung ‘uber die mechanische Erhlarung irrever-
sibler Vorgange’“, Ann. Physik 60 (1897) 392 (see also ref. 19, vol. 111, pp. 539-586).
38) Boltzmann, L., “0 ber einen mechanischen Satz H. Poincares”, Wiener Ber. 106 ( 1897) 12
(see also ref. 19, vol. III, pp. 5X7-595).
39) Boltzmann, L., “iiber eine von Hrn. Bartoli entdeckte Beziehung der Warmestrahlung zum
zweiten Hauptsatz”, Ann. Physik 22 (1884) 31 (see also ref. 19, vol. III, pp. 110~117).
40) Planck, M., “iiber irreversible Strahlunpsvorgange”, Berlin. Ber. (1895), (1896), (1897).
41) Boltzmann, L., “iiber irreversible Strahlungsvorgange I”, Berlin. Ber. (1897) 660-662 (see
also ref. 19, vol. III, pp. 615-617).
42) Boltzmann, L.. “iibcr irreversible Strahlungsvorgange II”, Berlin. Bet-. (1897) 1016-1018
(see also ref. 19, vol. 111, pp. 618~621).
43) Boltzmann, L., “iiber vermeintlich irreversible Strahlungsvorgange”, Berlin. Ber. (1898)
182-187 (see also ref. 19, vol. III, pp. 622-628).
44) Boltzmann, L., “iiber die sogenannte H-Kurve”, Math. Ann. 50 (1898) 325 (see also ref. 19,
vol. III, pp. 629-637).
45) Boltzmann, L., Math. Ann. 50 (1898) 637.
46) Boltzmann, L., “Zusammenhang zwischen den Satzen uber das Verhalten mehratomiger
Gasmolekiile mit Jacobis Prinzip des letzten Multiplikators”, 5 I of “Einige allgemeine Satze
tiber Warmegleichgewicht”, Wiener Ber. 63 (1871) 679 (see also ref. 19, vol. I, pp. 259-287
and ref. 46). Gibbs referred to this paper in ref. 6, p. viii.
47) Albert Einstein: Philosopher-Scientist, Library of Living Philosophers, Inc. (New York,
1949); Harper Torchbook edition (New York, 1959). “Autobiographical Notes”, p. 47.
48) Gibbs, J. W., Elementare Grundlagen der statistischen Mechanik, entwickelt besonders im
Hinblick auf eine rationelle Begriindung der Thermodynamik, J.A. Barth (Leipzig, 1905)
(German translation by E.ZermeIo).
49) Ehrenfest, P. and Ehrenfest, T., The Conceptual Foundations of the Statistical Approach in
Mechanics, originally published in German in the Encyklopgdie der mathematischen Wissen-
schaften, B.C. Teubner (Leipzig, 1912). English translation (by M.J. Moravcsik), Cornell
University Press (Ithaca, N.Y., 1959). See pp. 65-66 of the English translation.
50) See the remarks of Dugas, R., in La theorie physique au sens de Boltzmann, Griffon
(Neuchatel, 1959), especially pp. 267-269.
51) Hertz, P., “iiber die mechanischen Grundlagen der Thermodynamik”, Ann. Physik 33 (1910)
225, 537.
52) Einstein, A., “Bemerkungen zu den P.Hertzschen Arbeiten: ‘iiber die Grundlagen der
Thermodynamik’“, Ann. Physik 34 (1911) 175.
53) Tolman, R.C., Elements of Statistical Mechanics (Clarendon, Oxford, 1938).
EINSTEIN AND THE FOUNDATION OF STATISTICAL MECHANICS 477
54) Rosenthal, A., “Beweis der Unmoglichkeit ergodischer Gassysteme”, Ann. Physik 42 (1913)
796.
55) Plancherel, M., “Beweis der Unmijglichkeit ergodischer mechanischer Systeme”, Ann.
Physik 42 (1913) 1061.
56) Rosenthal, A., Ann. Physik 43 (1914) 894.
57) Fermi, E., 2. Phys. 24 (1923) 261.
58) Von Neumann, J., Proc. Nat. Acad. Sci. 18 (1932) 70, 263.
Birkhoff, G.D., Proc. Nat. Acad. Sci. 17 (1931) 650, 656.
Birkhoff, G.D. and Koopman, B.O., ibid. 18 (1932) 279.
59) Vibert Douglas, A., Forty Minutes with Einstein, J. Royal Astr. Sot. 50 (1956) 90.